Last updated 11 September 2017
Detect the guilty, protect the innocent
The sequence is familiar for TV crime series aficionados. Abby, the NCIS crime lab examiner, activates her powerful computer. She inputs a latent fingerprint collected from a crime scene, and instantly gets a portrait of the bad guy, with full identifying details.
She is doing an automatic
fingerprint identification, a process that automatically matches one or several unknown fingerprints against a database of known and/or unknown prints.
Predictably, reality is very different. The examiner needs to perform a quality check, plot minutia and then search...
But it is true to say that since they first emerged in the 1980s,
Automated Fingerprint Identification Systems (AFIS) used for criminal identification have become central to the work of police and other law enforcement agencies around the world.
By dramatically increasing the potential for
successful identification of a suspect, these systems have fundamentally changed the way that authorities approach the investigation of a wide range of crimes and criminal activities.
In this article, we will focus on 5 topics:
- The basics: biometrics for identification
- The technical challenges of AFIS
- The early stages: proving the value of AFIS
- The building blocks: a comprehensive investigation tool kit
- AFIS and the future of law enforcement
Let's dig in.
AFIS - 5 decades of research and development
At first glance, the principle of using modern technology to automate the laborious and time-consuming task of manually processing fingerprints taken from a suspect and/or crime scene appears straightforward.
However, the evolution of the
AFIS into a highly efficient and effective tool, capable of scrutinizing vast databases and providing potential fingerprint matches in a matter of minutes, is the product of intensive research and development that now stretches back over five decades.
And the process is ongoing.
Alongside the ever-present 'traditional' crime, the onset of new challenges, such as global terrorism and illegal immigration, have only heightened the need for authorities to identify individuals that might pose a threat to homeland safety and security.
At the same time, new biometric technologies - including iris and facial recognition – mean that the AFIS is rapidly transforming into the ABIS (Automated
Biometric Identification System), providing law enforcement agencies with an even more powerful tool.
With the new generation of ABIS software, fingerprint examiners can process multiple complex biometric transactions with high accuracy and link face recognition to fingerprint or to iris recognition. (source: Gemalto CABIS 7.0)
ABIS can process multiple complex biometric transactions with high speed and accuracy and link, for example, face recognition to fingerprint or to iris recognition, overcoming identification limitations commonly encountered in unimodal systems.
Fingerprints, biometrics and identification
Biometric identification is based on the principle that each individual can have a set of recognizable and verifiable data, which are unique and specific to them. For fingerprints, according to Sir
Francis Galton (Charles Darwin's cousin), the probability of finding two similar fingerprints is one in 64 billion even with twins.
Of course, the use of fingerprints as a means of identifying and convicting offenders has a history that stretches back well beyond the AFIS.
Today's livescan tenprint and palm print scanners are replacing the traditional ink based fingerprinting in many countries (source Gemalto)
Criminal identification systems originally emerged in the late 19thcentury. They were triggered by the landmark development of the Henry System of fingerprint classification in which fingerprints are sorted by physiological characteristics and anthropometrics also known as Bertillon system, in which measurements are obtained from suspects and filed.
In the UK, the Metropolitan Police started the use of biometrics for identification in 1901. In the US, it was initiated by the New York police in 1902 with French police initiating the same process in late 1902.
By the 1920s, the FBI had created its first Identification Department, establishing a central repository of criminal identification data for US law enforcement agencies.
All needed to be classified manually by an ever-growing team of staff. Similarly laborious manual searches had to be undertaken every time a potential match was sought.
But that's just part of the story…
AFIS to answer rising crime
The AFIS itself can trace its roots back to the electronics revolution of the 1960s.
The arrival of computers coincided with widespread concern over rising crime in the developed world.
In the US, a report compiled by the RAND Corporation proved particularly influential.
Significantly, it highlighted the opportunities for much more effective use of physical evidence – most notably fingerprints – to
improve crime solving performance.
Recognizing the potential of emerging technology to help achieve this goal, agencies including the FBI, UK Home Office, and police authorities in Japan and France all undertook research initiatives. Together this work helped to kick-start the development of the AFIS.
The challenges of automated fingerprint matching
The evolution of the modern AFIS required several more technological breakthroughs. And the scale of these challenges should not be underestimated.
Think about this for a moment.
To effectively replicate the work of skilled and experienced staff, a number of critical tasks had to be performed quickly, reliably and accurately.
Specifically these included:
- reading and capturing the traditional ink-on-card fingerprint image;
- detecting the 'minutiae' (distinguishing features) in the captured image;
- indexing the records;
- and comparing one set of minutiae data (taken from a suspect or crime scene, for example) with a large database of similar records.
Furthermore, the system needed to encompass both 'tenprints' and latent prints.
Tenprints or known prints
As the name suggests, tenprints comprises a complete set of fingerprints taken from an individual and collected on a single sheet. They are also referred to as "known prints" because the identity of the source of the impression is known.
Traditionally this has been done by applying a thin coat of ink across the ends of the fingers, then rolling them across a card. More recently, electronic 'livescan' devices have increasingly been used instead.
Latent palm print being lifted from a crime scene
Latent prints, in contrast, are those recovered from a crime scene or physical evidence, using chemical, physical and/or lighting techniques. Inevitably, these are often partial or highly fragmented, posing real problems in terms of reliable automated matching.
But let's see how a ten-fold increase in identification in latent prints in San Francisco changed the landscape for good.
From crime scene to courtroom
Once the key technical issues had been addressed, the AFIS needed to prove its value in the real world.
In this respect, a system supplied to the authorities in San Francisco in 1984 proved particularly significant. Notably, the city's new AFIS was part of a completely new 'crime scene to courtroom' philosophy.
This included the creation of a
dedicated crime scene investigation team, specially trained and equipped with its own labs and vehicles.
The impact was dramatic and widely publicized, and included
a ten-fold increase in identification of latent prints and swift decrease in burglary rates.
The use of an AFIS, and a much more focused approach to the collection and analysis of physical evidence, was totally justified. It became a must-have for large jurisdictions across the United States.
By 1999, 500 AFISes were deployed around the world.
Today, according to a 2017 study from
Markets and Markets, the automated fingerprint identification system market size is estimated to reach USD 8.49 Billion by 2020, at an estimated CAGR of 21.0% between 2015 and 2020. The major players include Gemalto Cogent, OT-Morpho and NEC among others.
A comprehensive investigative toolkit
The rapid adoption of AFIS inevitably led to further investment in development - a process that still shows no signs of abating. Consequently, the typical modern AFIS can perform tasks that include:
- Searching a known tenprint against a tenprint database
- Searching a latent print against a tenprint database
- Searching a latent against a latent database
- Searching a new tenprint against 'unsolved' latents
Further enhancements include the introduction of palm prints, interfacing the AFIS with other criminal justice information systems, interfacing with digital mugshots and livescan devices, and the use of multi-modal biometrics (e.g. facial, iris).
The human element is still critical
The process of quickly and reliably finding potential matches in huge databases requires vast computational power. Success depends on a wide variety of factors, notably the clarity of images and the degree of correspondence between the search print and the database print. Success rates of 30% are typical for latents - and remain highly dependent on the
skills of forensic technicians.
The technicians must know what to look for, and knowing what to look for takes 12 to 18 months of intensive training.
minutiae features are likely to be reviewed manually before a decision is made as to which one to focus a search on. Several searches may be needed using different parameters before a match is found. In the case of latents, it is also probable that several potential matches will be retrieved, requiring further analysis and interpretation by experts before a conclusion can be reached.
Automated Biometric Identification Systems provide improved efficiency with specialized workflows, such as the Gemalto CABIS 7.0 Unknown Latent Workflow.
A latent image is marked up and submitted for sequential search to both the latent fingerprint database (LFP) and the latent palm print database (LPP) thereby removing the need to remark and resubmit a second time.
Algorithms - the heart of a successful search
Effective use of highly sophisticated algorithms is a crucial element of the process. Over the years, many such algorithms have been developed, and enhanced continuously on the basis of real world experience. Commonly used examples include:
As the name suggests, image enhancement algorithms address the numerous issues that can affect the basic quality of latent or tenprint images.
Feature extraction algorithms are designed to identify the minutiae points (usually ridge endings and ridge bifurcations) that distinguish one print from another. These might also be supported by algorithms that can identify non-minutiae points, such as pores or textures. Indeed the combination of both minutiae and non-minutiae algorithms can prove particularly powerful in the search for a match.
Automatic indexing of fingerprints limits the sheer volume of data that an AFIS needs to process when searching for a match, significantly reducing the time taken to complete the task.
The design and choice of matching algorithms employed by the AFIS – and its operators - has a major impact on the number of potential matches, false positives and false negatives generated. Algorithms are also employed by an AFIS to provide a 'matching score'. This reflects the confidence with which a set of prints can be regarded as matching another found in the database.
Interconnection: IAFIS, IDENT and EURODAC
As the number of AFIS applications grew rapidly from the 1980s onwards, so did the need for integration and co-operation.
Established in 1999, the IAFIS (Integrated Automated Fingerprint Identification System) now upgraded to the
Next Generation Identification (NGI) is the world's largest collection of criminal history. Maintained by the FBI Criminal Justice Information Service, it contains the fingerprints of more than 118 million criminal and civil individuals as of June 2017 according to the FBI monthly fact sheet.
Cold case – a 1969 crime solved after almost 34 years in Houston, Texas.
Notably, the system architecture is designed to enable local, state, federal and international law enforcement communities, as well as civil organizations, to efficiently access or exchange critical information round the clock, 365 days a year.
In the United States, biometrics is also used to detect and prevent illegal entry into the country, grant and administer proper immigration benefits and facilitate legitimate travel and trade.
The Department of Homeland Security provides biometric identification services through its Office of Biometric Identity Management (OBIM), which supplies the technology for matching, storing, and sharing biometric data.
The system, called the Automated Biometric Ident
ification System or IDENT
, holds more than 200 million unique identities and processes more than 300,000 biometric transactions per day.
IDENT is the largest biometric repository in the U.S. government and shares critical biometric data with the Department of Defense and the Department of Justice to support homeland security, defense and justice missions.
Similarly, in Europe, the Eurodac biometric system (European Dactyloscopy System) is the largest multi-jurisdictional AFIS in the world serving 32 countries. Eurodac's is the EU's asylum fingerprint database and contains the fingerprints of all asylum applicants from each Member State, as well as fingerprints from persons apprehended in an irregular border crossing.
Budapest, Hungary – September 2015: Refugees standing up and waiting to take a train to Austria. The Eurodac system enables the comparison of fingerprints of asylum applicants and persons apprehended in connection to an irregular or illegal border crossing. 28 EU Member States and 4 Associated Dublin States (Iceland, Norway, Liechtenstein and Switzerland) use the system.
Shaping the future of law enforcement
We've seen that over the space of several decades, the AFIS has made huge strides.
But they are still far from a 'magic bullet'. Certainly they cannot replicate completely the complex analytical skills of a forensic expert.
However, the sheer speed and accuracy with which a modern AFIS can now work allows these skills to be deployed with the highest possible degree of efficiency.
And going forward, the role of the AFIS is only likely to grow in significance. Ultimately, the introduction of AI (Artificial Intelligence) may deliver another step change in performance.
More immediately, a much richer array of biometric data is being embraced, further increasing the chances of matching physical evidence, or a crime scene, to a suspect. At the same time, this is raising new ethical questions, and decisions will need to be made in terms of how an individual's data is collected, stored, shared and used.
Different societies will reach different conclusions, but there can be little doubt that both AFIS and ABIS will provide invaluable support for a wide range of law enforcement teams, as they seek to build safer communities and bring to justice all those who pose a threat to the security and well-being of law-abiding citizens.
Visit our July 2017 web dossier to know more on
biometric data protection legal frameworks in Europe and the United States.
If you want to know more on AFIS history, we suggest this
remarkable document from Kenneth R. Moses.
We help forensic examiners find answers fast
Leveraging over 27 years of biometric experience from trusted industry leader, Cogent, Gemalto
Cogent Automated Biometric Identification System (CABIS) provides the multi-biometric tool that can help examiners find answers quickly and efficiently.
Gemalto CABIS is used for investigation, identification and verification in civil and border identification and law enforcement applications. Cogent AFISes are already deployed to over 200 applications such as the IDENT or EURODAC systems, in more than 80 countries worldwide.
Now it's your turn
If you've something to say on AFIS/ABIS, a question to ask, or have simply found this article useful, please leave a comment in the box below. We'd also welcome any suggestions on how it could be improved, or proposals for future articles.
We look forward to hearing from you.